I met Alexa’s dad last week. Ok, I met the bloke behind the technology that powers Amazon’s smart speaker, William Tunstall-Pedoe.
He was speaking at CogX19, The Festival of AI and Emerging Technology, where a global smörgåsbord of tech geeks gave insights into the good, the bad and the ugly of Artificial Intelligence, chatbots and more.
I went because I’m paranoid. The idea that AI could snaffle up our writing business causes me sleepless nights. Creative writing, customer conversations, even, god forbid, headline writing, are all in the robotic cross-hairs.
Tunstall-Pedoe is a self-confessed AI speech nerd. His business, True Knowledge, pioneered some pretty amazing voice recognition technology. Amazon saw the potential, bought the business for $26m, and the rest is history.
Or rather, the future: smart speakers like Alexa have the fastest rate of uptake of any technology in recent decades. Faster than smartphones, TVs, radios, the internet, or computers. Sales of smart speakers are estimated to reach 1 billion by 2025. Voice is here to stay.
However, interacting fluently with natural speech is still a stretch for voice technology. There are over 80 natural language platforms at the moment and none can decipher vernacular speech that well – for now.
In the meantime, the rise of the chatbot continues apace. We’re getting increasingly used to interactions with automated response systems, on screen or through Alexa etc. This more rule-based, guided version of natural speech is known as ‘conversation design’ – and it’s going to play a big role in brand marketing.
So, conversation, that’s an opportunity for writers, right? We’re always telling people to write like they speak, so this should be a cinch. Or will robots do it all for us?
The AI elephant in the room
Let’s start by addressing the biggie: how creative is AI? Is our role as headline writers under threat? Is our expertise in tone of voice unwanted?
At the Open AI project in San Fransisco, they’ve been developing AI’s potential to generate music – not rule-based algorithms, but genuine, creative compositions. Open AI define the stages of creative AI as:
- Mimicry: matching human styles
- Combining existing styles in novel ways
- Inventing an entirely new style
- Transcending the limits of what we previously defined as possible
They’ve got to stage 2 – bringing Chopin and Bon Jovi together made interesting listening!
Mimicry is already happening in copywriting. Subject line generators like Phrasee analyse and write subject lines so quickly that they hone in on the lines that have the best open rates and conversion.
Because automated systems measure and check far faster and more effectively than a human could, Phrasee can create a more successful subject line strategy. But it won’t be having a chat with you anytime soon. Given the right input, AI can create convincingly bland copy too (shudders). And chatbots are rapidly improving their ability to answer our questions based on predicted responses.
The language project at Open AI shows what an intelligent system is capable of writing – here’s their piece on unicorns in the Andes. They’re so concerned about the ‘malicious application of the technology’ that they won’t release the full model. Scary.
Step two offers interesting opportunities for brands. In the same way you can brief musical AI systems to play or combine tracks in different styles, brands could adapt messages or content for tone depending on your mood or time of day. Consumers will be able to use what’s known as ‘gestural AI’ to change the mood or style of messages.
Step three, an entirely new style, is where AI creativity begins to falter. Would an AI writer call a Volkswagen Beetle ‘Lemon’? Or Pot Noodle ‘The slag of all snacks’? Bit of a leap at the moment. And as for step four …
Crucially, where AI and human creativity part company is that humans crave a back story. Even if AI can compose or write amazing music or poetry or headlines, what’s the point? We love the stories behind artists that inform our sense of the composition. So do brands. Think Apple and Steve Jobs, the Innocent founders’ story, or Levi Roots and his Reggae Reggae sauce. AI creativity will always be synthetic and can’t make up for a lack of real story.
How personal do you want your personalisation?
Good writing is about empathy: writers have the ability to project, to tap into what an audience is feeling. We make our copy engaging by making it capture a mood. AI behavioural recognition technology is clearly a threat to this particular skill set. AI will help brands identify moods or situations and adapt accordingly. Devices will get better and better at knowing what we’re likely to ask, or need, at any particular time of day. Ultrapersonalisaton was the word of the day. At times CogX all got a bit Black Mirror-y – check out this vision of AI and your daily life from Constellation AI.
The win for copywriters is that a higher degree of personalisation, particularly with more targeted media, is that consumers will expect a higher standard of relevant content – and that still can’t be created by AI writing.
Let’s chat chatbots and tone of voice
Pandorabots, the world’s biggest bot company, estimate that up to 30% of enquiries to chatbots are inappropriate – sex, drugs, or abuse – chatbots need to know when to stop trying to answer. Knowing their limitations is crucial.
Humanise AI have created Gem, a chatbot for hotels. It’s pretty slick and can handle most enquiries. And it knows when to hand over to a human.
Even if the questions aren’t inappropriate, the chatbot may not understand what it’s been asked and will need to hand over. To segue successfully between automated responses and real people depends on having a consistent tone of voice. This is where copywriters can work with brands to make sure their conversation design embodies all the values in their tone of voice. A consistently branded voice is central to a good customer experience.
For a good example of a chatbot with tone, check out Mitsuku, created by Steve Worswick in Leeds (have a chat and you’ll see why Leeds is relevant).
Chatbots represent an exciting future for brands and copywriters. Copywriters can help shape the brand voice that conversation design brings to life in a chatbot or more sophisticated voice technology. Customer service teams will still need to invest in tone of voice training so that their voice is consistent with the automated version where it hands over to a human handler.
And until an AI copywriter creates a headline with the elegant simplicity and wit of The Economist ads, or a tone of voice as quirky as Innocent, I think we can all sleep easy.
This post was written by Richard Spencer